Advanced Statistical Methods for Brain Connectivity Analysis Restricted; Files Only
Ran, Jialu (Summer 2024)
Abstract
The analysis of brain network connectivity has become an important tool for investigating brain function and various disease characteristics. However, this type of analysis is challenging due to the complex structure and high dimensionality of brain imaging data. To extract meaningful information from these data, sophisticated analytical methods are required. This dissertation focuses on the development of statistical methods to enhance our understanding of the human brain.
In Chapter 1, we introduce the background of brain connectivity analysis, discuss the associated challenges, review existing solutions and their drawbacks, and highlight our contributions to the field.
In Chapter 2, we introduce dyna-LOCUS, a method for identifying neural circuits in dynamic brain functional connectivity. This method performs dynamic connectome source separation using low-rank modeling of connectivity traits with sparsity and temporal smoothness regularization. We validate the performance of dyna-LOCUS through simulations and apply it to the Philadelphia Neurodevelopmental Cohort (PNC) study. This application identifies latent neural circuits underlying observed dynamic connectivity, reveals key brain regions that drive each of these circuits, investigates their temporal expression profiles, and identifies neural circuits associated with gender and neurodevelopment.
In Chapter 3, we develop a longitudinal-LOCUS method to study changes in brain connectomes over time. This method decomposes longitudinal functional connectivity (FC) measures using blind source separation with low-rank structures and angle-based sparsity regularization. We present an efficient iterative node-rotation algorithm to solve the non-convex optimization problem for learning longitudinal-LOCUS. Simulations demonstrate superior accuracy in recovering latent sources and mixing coefficients compared to existing methods. We apply the method to the Adolescent Brain Cognitive Development (ABCD) data to investigate developmental changes in neural circuits and their differences between genders.
In Chapter 4, we address a challenge that arises in brain imaging data when participants move their heads during a brain scan. This motion can cause spurious associations when studying differences between functional connectivity across groups. We propose decomposing neural and motion-induced sources of group differences under a causal mediation framework. We establish the theoretical properties of our proposed estimators and validate the theory using simulation studies. The framework is applied to estimate the difference in functional connectivity between autism spectrum disorder (ASD) children and typically developing children in the ABIDE study. Our analyses indicate that some long-range connections between a seed region in the default mode network and frontal-parietal regions exhibit hyperconnectivity in ASD. Naively including high-motion children appears to cause spurious connectivity differences. Naively excluding high-motion children removed group differences.
In Chapter 5, we expand on the second project by developing MoCo, an R package for motion control in MRI studies. In addition to motion control, MoCo allows for missingness in MRI data due to collection or preprocessing issues. We demonstrate MoCo with examples related to ASD and attention-deficit/hyperactivity disorder (ADHD), showing it effectively mitigates motion artifacts, enhances data utilization, addresses selection biases, and is more robust to preprocessing pipelines.
Table of Contents
1 Introduction 1
2 Unveiling Hidden Sources of Dynamic Functional Connectome through a Novel Regularized Blind Source Separation Approach 8
2.1 Introduction................................ 8 2.2 MaterialsandMethods.......................... 14 2.3 Simulation................................. 29 2.4 Investigating dynamic functional connectome for the Philadelphia Neurodevelopmental Cohort(PNC) study.................. 32 2.5 Appendix ................................. 533 Investigating latent neurocircuitry traits underlying longitudinal brain functional connectome 59
3.1 Introduction................................ 59 3.2 Material and Methods .......................... 63 3.3 Estimation Algorithm and Selection of Tuning Parameters . . . . . . 66 3.4 Simulation................................. 72 3.5 Data analysis of longitudinal functional connectivity in ABCD study 774 Nonparametric Motion Control in Functional Connectivity Studies in Children with Autism Spectrum Disorder 88
4.1 Introduction................................ 88 4.2 Methods.................................. 91 4.3 EstimationandInference......................... 99 4.4 Simulation study ............................. 109 4.5 Data analysis of functional connectivity in ASD . . . . . . . . . . . . 1115 MoCo: A package for removing motion artifacts in brain phenotype analysis 117
5.1 Introduction................................ 117 5.2 Background and theory.......................... 120 5.3 Tutorial .................................. 126 5.4 Dataset and analysis ........................... 1366 Conclusion and Future Work 145
7 Appendix 154
Bibliography 223
About this Dissertation
| School | |
|---|---|
| Department | |
| Degree | |
| Submission | |
| Language |
|
| Research Field | |
| Keyword | |
| Committee Chair / Thesis Advisor | |
| Committee Members |
Primary PDF
| Thumbnail | Title | Date Uploaded | Actions |
|---|---|---|---|
|
|
File download under embargo until 21 August 2025 | 2024-08-16 20:07:15 -0400 | File download under embargo until 21 August 2025 |
Supplemental Files
| Thumbnail | Title | Date Uploaded | Actions |
|---|